I have a model Post:
class Post(models.Model):
post_title = models.CharField(max_length=120)
post_subtitle = models.TextField()
post_text = models.TextField()
vector_column = SearchVectorField(null=True)
class Meta:
indexes = (GinIndex(fields=['vector_column']),)
and I've created a trigger in the database to update the vector_column
value:
create function core_post_trigger() returns trigger as $$
begin
new.vector_column :=
setweight(to_tsvector('pg_catalog.english', coalesce(new.post_title, '')), 'A') ||
setweight(to_tsvector('pg_catalog.english', coalesce(new.post_subtitle, '')), 'B') ||
setweight(to_tsvector('pg_catalog.english', coalesce(new.post_text, '')), 'C');
return new;
end
$$ language plpgsql;
create trigger vector_column_trigger
before insert or update on core_post
for each row execute procedure
core_post_trigger();
And I search through this model like this:
Post.objects.filter(vector_column=SearchQuery(query, config='english', search_type='websearch')
Although I've applied weights in this search, I did not apply ranking (to_tsrank
). I know that I can apply rank in Django like this:
vector = SearchVector('post_title', weight='A', config='english') + SearchVector('post_subtitle', weight='B', config='english') +
SearchVector('post_text', weight='C', config='english')
query = SearchQuery(query, search_type='websearch')
Post.objects.annotate(rank=SearchRank(vector, query)).order_by('-rank')
The question is that I don't know how to use SearchRank
with my vector_column
column, as this column has a GinIndex. If I use this vector
variable created above, I will not use the index, and then this search will be slow. I've tried to write this query in various ways, but I don't know how to do. I tried, for example, this:
Post.objects.annotate(SearchRank(vector_column, SearchQuery('Angel', config='english'))).order_by('-rank')
But it throws an error vector_column is not defined
.
How can I achieve it?
I found it. I needed to use a F expression:
Post.objects.annotate(
rank=SearchRank(F('vector_column'),
SearchQuery(query, config='english'))
).order_by('-rank'), 15